[kixelated]

Problem

quest/m0/ci/slow-tests.md: the 103 test IDs frozen in moq-dev/moq#6914 took 137.83 CPU-seconds at baseline. The target is under 68.92 s, with every case and its coverage kept. Current main measures 100.55 s.

Result

Two full normal runs of this branch measure the cohort at 61.02 s and 67.49 s. Both are under 68.92 s. All 103 IDs ran and passed. Whole-suite test CPU fell from 206.2 s to 166.0 to 178.5 s.

Run-to-run noise is about 10%, mostly in the export timing tests under host contention. Run 2 clears the target by only 1.4 s, so a noisier host could put it over.

Module	Tests	Main s	Run 1 s	Run 2 s
moq-mux container::ts::export_timing_test	14	24.7	11.2	15.3
moq-mux container::ts::recovery_test	18	11.8	10.2	10.5
moq-net::integration routing_contracts	5	10.9	6.3	6.5
moq-quic	9	8.4	3.0	3.4
moq-net::integration mesh_withdraw	6	8.4	8.2	8.8
moq-mux jitter::tests	8	6.8	5.1	5.1
moq-json	8	6.4	1.8	2.1
moq-net	9	6.0	3.4	3.5
moq-stats	4	4.5	4.4	4.4
moq-net::integration catalog_churn	2	2.0	1.9	2.0
moq-tokio::integration	4	1.7	1.4	1.2
moq-audio	2	1.5	0.7	1.0
qmux	1	1.3	0.4	0.4
moq-video	1	1.2	0.2	0.2
moq-mux rate::tests::model	1	1.2	0.2	0.2
moq-pattern::integration	1	1.1	0.6	0.6
everything else	10	2.9	2.1	2.4

Changes

No test was deleted, ignored, or moved to nightly. No assertion or bound was loosened.

moq-mux

• export_timing_test: the Live harness advances its own kio::clock::Deadlines queue instead of sleeping on the tokio paused clock, the same switch the jitter tests made in test(mux): reduce fixture CPU and checkpoint frozen cohort #6914.
  • A leg that reads after every frame now polls only once its export has been woken. 41k of 60k polls had nothing to wake them, and none returned a slot.
  • assert_same_packets compares 188 bytes at a time and only drops to bytes to name the offset that differs.
  • Every rendered slot's payload hash matched before and after across all 21 tests.
  • drain advances that queue 10 ms itself when nothing is due, instead of a kio::clock::timeout armed on a queue nothing else advances, which hung four tests once main's export stopped reporting the end there.
  • Frames now arrive on their exact microsecond. Tokio rounded sleeps up to the millisecond, so some slot send instants moved by up to about 21 ms.
• export_timing_test: the three drifting-source cases measure 60 s of media instead of 300 s. See Decisions.
• recovery_test:
  • x265 runs with pools=none:frame-threads=1, since it ignores -threads 1. This also makes the bitstream the same on every host.
  • The reference decode hashes pictures with murmur3 instead of MD5.
  • A run concatenates its TS output once instead of up to five times.
• jitter: the drift and feed helpers drain due frames without collecting them. Steer::measure (production) skips the depth sort when only one track qualifies, because the depth only breaks ties between tracks.
  • The jitter_steering release bench: 1-track cells about 7% faster (1/300: 287 to 267 µs), 4, 16 and 64-track cells unchanged within noise.
• rate model proptest uses RngAlgorithm::XorShift. ChaCha seeding on every weighted prop_oneof fork was 60 to 80% of the run. The case count, seed and strategies are unchanged.
• fixture::Mux::packet builds into a [u8; 188], with identical bytes.

moq-net

• routing_contracts reference model (test only): each identity's route is selected once per pass instead of up to seven times. Links live in a dense matrix, and per-peer views are restricted once per distinct interest.
  • handoff::loop_free skips identities that were never originated.
  • A temporary hash of every Net's full state, events, stats and queues was byte-identical before and after across all 56 routing_contracts tests.
• front::tests::bounded_sequences_hold_the_invariants renders Debug only when an event emitted something other than Detach. It is the same check in contrapositive form, with cheap fields compared first.
• a_busy_track_yields_to_its_siblings and a_busy_subscription_yields_to_its_siblings serve 2,000 groups instead of 10,000. See Decisions.

moq-json

• The window encoder reads its group budget from a private max_group_bytes field, and the stream encoder its per-record limit from a private max_frame_bytes field (main's feat(net)!: shed group fronts past a 32 MB tail or under pool debt #6993 dropped the stream's group budget). Both default to MAX_CACHE_BYTES, so behavior is unchanged. This mirrors the snapshot encoder's existing field.
  • Their oversize tests use a 32 KiB limit and exercise the same refusal and roll branches. a_compressed_log_is_charged_its_inflated_size from the table below no longer exists on main.
• The two codec property tests use XorShift, for the same reason as rate.

moq-quic

• The sim payload() returns a slice of one shared 251-byte cycle instead of building 64 KiB byte by byte on every send attempt. Streams under 251 apart still differ at every offset.
• subscribe() reads RUST_LOG and defaults to info instead of formatting every TRACE event. The per-step client and server spans are debug. The runs are deterministic, so a failure replays with RUST_LOG=trace. Main's subscribe_at is gone: its one caller, the probe runs, wanted info, which subscribe() now gives.

Elsewhere

• qmux keep-alive: builds its 100 datagrams once instead of 200,000 times.
• moq-audio decode corpus: replays 8 kHz PCM, since only a packet's duration feeds the frame term.
• moq-video transcode: frames are an eighth as wide. Color::infer reads only the height.
• moq-cli passthrough: joins its output with concat().
• moq-relay mesh: reuses one reqwest::Client per mesh. Building one per 10 ms poll loaded the system trust roots each time.
• moq-tokio zero-backoff: a 100 ms give-up window instead of 500 ms. The bug resets the window on every session, so any length catches it.
• moq-hls (production): a file playlist builds no HTTP client, so it also skips the rind rustls install main added before building one. same_origin already kept every fetch on the entry's scheme, so it never used one, but building it parsed the system trust roots.
• Cargo.toml: rubato builds at opt-level 3 in dev, as sonora already does. Its SIMD intrinsics are the whole resampler invariant test (0.9 to 0.14 s).

Decisions

Settled by the maintainer in a /quest-complete session, which accepted both recommendations.

1. Measure the drifting-source cases over 60 s instead of 300 s (a_slow_source_clock_keeps_the_delay, a_fast_source_clock_keeps_the_delay, a_source_past_the_limit_is_counted; commit test(mux): measure a drifting source's rate over a minute). It saves about 6.5 s, without which the cohort misses the target.
   • Measured rates at 10 s are within 0.002 ppm of the 300 s values (29.99919 vs 29.99936; -30.00099 vs -30.00116; 399.84023 vs 399.84032), against bounds of 0.5 and 1 ppm. Neither RATE_WINDOW nor RESPONSE (both 600 s) fills by 300 s, so the steering stays in the same regime.
   • Margins at 60 s and 300 s: spread 15.7 vs 22.8 ms under a 187 ms bound, 0 drops, margin 498 vs 491 ms.
   • Tradeoff: 30 steering steps instead of 150, so fewer chances for the spread check to catch an anchor glitch.
   • Settled: keep.
2. Serve 2,000 groups instead of 10,000 in the two busy-sibling yield tests (saves about 1.0 s).
   • Every poll is still checked against the 32-group budget, the sibling's turn and the stall bound. Two producer batches of 1,000 keep the backlog well past the budget, and the original bug shows on the first poll.
   • Settled: keep.

Also chosen while iterating:

• XorShift draws different concrete proptest cases than ChaCha did, from the same distribution with the same seed and case count.
• moq-quic test logs default to info.
• The rubato opt-level follows the sonora precedent.

Impact

• Public API: none. The moq-json budget field and the moq-hls client are private.
• Wire: none.
• Production behavior: unchanged. Steer::measure picks the same slope, and a file HLS import makes no HTTP request, as before.

Validation

• quest gate nix develop -c just ci origin/main on cf952d0 (main 3164bce): every module passed but check flake, which fails on this host because the Nix builder cannot create user namespaces (unshare: write failed /proc/self/uid_map). test rs: 10,908 tests passed, 124 skipped.
• Earlier: just check passed at 6d9931e (10,386 tests).
• quest check --base origin/main passed.
• Measured with the wrapper and command from test(mux): reduce fixture CPU and checkpoint frozen cohort #6914 (wrapper SHA-256 d16e86fb379a8d68b301f1471d5f135517faa5225f8fbaa1156f3e7405e5ef48), under just rs test --workspace --exclude moq-net-fuzz --exclude moq-quic-fuzz, zero retries. Same machine as the frozen baseline (Intel Core Ultra 7 270K Plus, 24 logical CPUs).
  • Main is fbeac8a9fd.
  • Runs 1 and 2 are this branch at 6d9931ef90; the later commit only removes the quest.
  • The host was shared, with a load average of about 10 during the runs.

Tried and dropped

• select_nth_unstable_by for the jitter medians: the release bench regressed 15 to 25% at 300 floors. sort_unstable_by was neutral in release and cost 45% more in debug.
• opt-level = 3 for serde_json: about 20% on the stats oversize tests, because its generic code is instantiated in the calling crates at opt-level 0.
• Splitting the moq-net mock's Link::route by stream kind: at most 4% of the mesh tests, and it changes the wakeup order the pinned mesh counts depend on.
• -aac_coder fast for the recovery clips: no change.

Follow-ups

• Encode the recovery clips once per run instead of per test process. Each test pays an FFmpeg encode plus a reference decode, roughly 35 to 40% of recovery_test's CPU. Options: commit the clips (about 320 KB each), generate them in a nextest setup script, or keep a lock-guarded cache keyed by FFmpeg version.
• Cheaper un-woken Export::poll_next. Track fill through moq-net's Subscriber::poll_group is about 24% of an export poll, and catalog polling through moq_json::snapshot::Consumer::poll_latest into Cursor::poll_next_group is 11 to 19%. This helps every export test and production. Needs a benchmark.
• moq-net session cost in the random-mesh tests. AnnouncePrefix::poll_run, origin sync_route/requests_route, OriginScope::relative rebuilding Patterns on every call, and per-session setup (about 25%) dominate mesh_withdraw. Needs benchmarks.
• moq-quic frame::Iter yields one Frame::Padding per padding byte. In debug, that move is about 30% of handshake CPU. Collapse runs, after checking what the frame_rx padding stats should count.
• tracing-test's no-env-filter, enabled by moq-tokio, unifies across the build, so every crate formats at TRACE. That is 11 to 17% of the moq-tokio integration tests.
• moq-stats oversize tests (3.3 s) push past the hard 32 MiB cache through a debug JSON pipeline. Production option: the snapshot encoder refuses an oversized plain snapshot before parsing it back, as it already does for deflate.
• a_fill_refused_group_by_group_stays_linear (1.8 s) runs 100,000 refusals. A smaller count catches a plain quadratic but not a cheap per-refusal copy.

Per-test CPU seconds: frozen baseline, main, and two runs of this branch

Binary ID	Frozen test ID	Baseline	Main	Run 1	Run 2
moq-mux	container::ts::export_timing_test::two_legs_render_a_drifting_source_the_same_way	5.64	5.81	3.77	4.47
moq-net::integration	routing_contracts::handoff::sweep_handoff	4.51	4.38	2.61	2.64
moq-mux	jitter::tests::the_clock_follows_a_drifting_source_for_a_day	4.94	3.59	2.53	2.53
moq-mux	container::ts::export_timing_test::a_joiner_keeps_the_system_clock_in_tolerance	3.48	3.48	2.33	3.50
moq-mux	container::ts::export_timing_test::a_fast_source_clock_keeps_the_delay	3.71	3.47	0.46	0.66
moq-mux	container::ts::export_timing_test::a_source_past_the_limit_is_counted	3.18	3.43	0.49	0.78
moq-mux	container::ts::export_timing_test::a_slow_source_clock_keeps_the_delay	3.15	3.43	0.46	0.53
moq-net::integration	routing_contracts::handoff::sweep_requests_reach_a_quiet_origin	3.40	3.26	1.83	1.88
moq-net::integration	mesh_withdraw::residual_random_meshes	3.32	3.08	3.00	3.22
moq-net::integration	mesh_withdraw::capped_random_meshes_hunt_less	3.33	3.01	2.96	3.17
moq-net	model::front::tests::bounded_sequences_hold_the_invariants	2.25	2.20	0.58	0.59
moq-quic	tests::sim::random_workloads	2.13	2.03	0.72	0.74
moq-net::integration	routing_contracts::sweep_forgetting_distances	2.06	1.99	1.10	1.14
moq-stats	produce::tests::an_oversize_group_leaves_other_groups_readable	2.26	1.92	1.89	1.91
moq-net	model::track::test::a_fill_refused_group_by_group_stays_linear	2.35	1.80	1.77	1.80
moq-quic	tests::sim::replays_from_seed	1.83	1.79	0.17	0.22
moq-json	window::encoder::test::a_header_larger_than_the_group_cache_is_rejected	1.37	1.39	0.00	0.00
moq-stats	produce::tests::oversized_frame_ends_both_flavors	2.08	1.35	1.35	1.36
qmux	session::tests::websocket_keep_alive_runs_during_continuous_output	1.35	1.30	0.37	0.37
moq-json	window::encoder::test::an_op_that_would_evict_the_header_rolls_first	1.27	1.28	0.00	0.00
moq-video	encode::encoder::tests::transcode_keeps_the_space_the_source_signalled	1.24	1.18	0.18	0.18
moq-mux	rate::tests::model::invariants_hold_on_random_walks	1.15	1.15	0.20	0.19
moq-net::integration	catalog_churn::catalog_churn_lite_07	1.14	1.13	1.11	1.12
moq-mux	container::ts::recovery_test::hevc_keeps_radl_after_a_hole	1.13	1.11	0.82	0.86
moq-pattern::integration	pattern_exhaustive::intersect_matches_exactly_what_both_match	1.09	1.09	0.57	0.57
moq-mux	container::ts::recovery_test::hevc_trims_only_orphaned_rasl	1.06	1.05	0.84	0.86
moq-quic	tests::probe::congested	0.97	0.99	0.39	0.56
moq-net::integration	routing_contracts::sweep_forwarding_progress	0.94	0.93	0.51	0.52
moq-stats	produce::tests::reheld_wire_presence_refreshes_two_relay_caches	1.70	0.90	0.86	0.87
moq-mux	container::ts::recovery_test::a_skip_across_a_resolution_change_takes_the_new_parameters	0.84	0.89	0.75	0.81
moq-json	window::properties::window_codec_property_matches_every_readers_logical_window	0.89	0.89	0.57	0.71
moq-mux	container::ts::recovery_test::a_lost_discontinuity_marker_still_breaks_the_clock	0.84	0.88	0.75	0.79
moq-json	snapshot::encoder::test::snapshot_codec_property_reconstructs_each_json_value	0.85	0.87	0.49	0.57
moq-audio	resample::tests::duration_and_samples_are_invariant_across_partitions_and_epochs	0.87	0.87	0.14	0.13
moq-mux	container::ts::recovery_test::a_declared_discontinuity_survives_loss_either_side	0.84	0.85	0.77	0.80
moq-net::integration	catalog_churn::catalog_churn_lite_06	0.83	0.83	0.80	0.83
moq-mux	container::ts::export_timing_test::frames_held_through_a_silence_drop_as_late	0.78	0.82	0.65	0.70
moq-quic	tests::bbr_marking_versus_dropping	0.73	0.78	0.28	0.30
moq-quic	tests::probe::borderline_skips_probe_rtt	0.68	0.74	0.30	0.31
moq-net::integration	mesh_withdraw::partial_mesh_withdraw_then_republish	0.84	0.73	0.76	0.79
moq-net	ietf::publisher::group_priority_test::a_busy_track_yields_to_its_siblings	0.73	0.72	0.15	0.16
moq-json	stream::test::an_oversized_record_is_refused_and_the_log_continues	0.64	0.69	0.00	0.00
moq-mux	container::ts::export_timing_test::a_skip_at_the_join_keeps_one_clock	0.87	0.69	0.48	0.82
moq-mux	container::ts::export_timing_test::a_skip_while_running_keeps_one_clock	0.66	0.68	0.48	0.77
moq-mux	jitter::tests::a_wandering_track_sent_latest_does_not_move_the_rate	1.14	0.67	0.63	0.63
moq-audio	decode::consumer::tests::the_decode_path_conforms_to_the_corpus	0.62	0.67	0.56	0.83
moq-quic	tests::probe::rises_to_target	0.64	0.62	0.27	0.29
moq-mux	jitter::tests::closing_a_late_start_is_not_charged_to_the_rate	0.94	0.61	0.39	0.39
moq-mux	container::ts::recovery_test::a_skipped_group_resumes_at_the_next	0.59	0.60	0.51	0.51
moq-mux	container::ts::recovery_test::a_skip_across_the_wrap_keeps_the_clock	0.56	0.60	0.51	0.52
moq-mux	container::ts::recovery_test::a_cut_group_keeps_what_it_delivered	0.60	0.59	0.49	0.52
moq-quic	tests::sim::priority_on_a_clean_path	0.62	0.59	0.45	0.50
moq-mux	container::ts::recovery_test::a_stalled_track_leaves_the_others_running	0.52	0.57	0.48	0.48
moq-net::integration	mesh_withdraw::full_mesh_withdraw_retracts_once_lite04	0.60	0.56	0.55	0.58
moq-mux	container::ts::recovery_test::skewed_audio_survives_a_video_skip	0.57	0.56	0.52	0.54
moq-mux	container::ts::export_timing_test::a_track_sent_later_than_the_delay_loses_nothing	0.51	0.56	0.40	0.56
moq-mux	container::ts::recovery_test::a_stall_drops_what_it_made_late	0.53	0.55	0.49	0.52
moq-mux	container::ts::recovery_test::a_clean_path_loses_nothing	0.56	0.54	0.51	0.51
moq-mux	container::ts::export_timing_test::a_cut_judges_held_video_when_the_loss_is_observed	0.89	0.54	0.39	0.63
moq-mux	container::ts::recovery_test::a_congested_path_sheds_what_it_made_late	0.55	0.54	0.49	0.51
moq-net::integration	mesh_withdraw::full_mesh_withdraw_retracts_once_lite06	0.61	0.53	0.52	0.58
moq-tokio::integration	broadcast::zero_initial_backoff_still_gives_up_on_a_flapping_peer	0.48	0.53	0.13	0.12
moq-mux	jitter::tests::a_late_start_within_the_drift_limit_never_fails	0.91	0.52	0.39	0.40
moq-mux	container::ts::recovery_test::open_gop_trims_only_the_orphans	0.51	0.52	0.49	0.51
moq-json	stream::test::a_compressed_log_is_charged_its_inflated_size	0.53	0.51	0.00	0.00
moq-mux	container::ts::recovery_test::a_path_stall_drops_what_it_made_late	0.52	0.50	0.46	0.48
moq-mux	container::ts::export_timing_test::the_video_held_at_the_end_goes_out	0.46	0.49	0.39	0.45
moq-mux	container::ts::recovery_test::a_late_catalog_starts_on_a_whole_group	0.50	0.49	0.43	0.43
moq-mux	container::ts::recovery_test::a_join_starts_on_a_whole_group	0.51	0.49	0.45	0.47
moq-net	lite::publisher::tests::a_busy_subscription_yields_to_its_siblings	0.46	0.48	0.11	0.12
moq-quic	tests::probe::overflow_keeps_probe_rtt	0.45	0.48	0.21	0.33
moq-net::integration	fetch_fanout::sequential_viewers_share_the_relays_fetched_groups	0.49	0.48	0.46	0.47
moq-mux	container::ts::recovery_test::the_graders_fail_what_they_should	0.42	0.45	0.43	0.44
moq-mux	jitter::tests::a_spell_on_the_steady_track_keeps_the_rate_on_it	0.77	0.45	0.45	0.44
moq-net::integration	mesh_withdraw::partial_mesh_withdraw_retracts_once	0.55	0.45	0.44	0.46
moq-tokio::integration	receive_credit::bufferbloat_is_not_a_stuck_window	0.56	0.45	0.56	0.41
moq-mux	container::ts::export_timing_test::a_successor_settles_held_video_across_a_skip_or_restart	0.41	0.45	0.35	0.52
moq-cli::bin/moq	publish::tests::ts_passthrough_crosses_a_relay_through_a_flagged_jump	0.49	0.44	0.38	0.40
moq-mux	container::ts::passthrough_test::a_long_gop_splits_at_the_byte_ceiling_between_objects	0.45	0.43	0.15	0.15
moq-mux	container::ts::export_timing_test::video_held_for_most_of_the_delay_drops_nothing	0.39	0.42	0.27	0.48
moq-quic	tests::bbr_burst_after_long_idle_keeps_learned_bandwidth	0.40	0.41	0.17	0.19
moq-mux	container::ts::export_timing_test::live_audio_shares_its_pes	0.37	0.40	0.31	0.45
moq-net::integration	routing_contracts::handoff::sweep_scoped_handoff	0.36	0.38	0.29	0.31
moq-mux	jitter::tests::a_step_in_queueing_does_not_move_the_estimate	0.46	0.38	0.21	0.21
moq-tokio::integration	broadcast::broadcast_rejoin_replays_a_current_warm_cache	1.05	0.38	0.36	0.38
moq-json	snapshot::encoder::test::memo_matches_the_value_diff	0.37	0.37	0.36	0.40
moq-json	snapshot::test::a_rejected_update_leaves_the_previous_value_readable	0.35	0.35	0.34	0.38
moq-cli::bin/moq	hls::test::an_ended_playlist_announces_its_catalog	0.52	0.34	0.09	0.09
moq-relay::integration	cluster_unknown::every_ingest_version_crosses_a_redundant_mesh	0.54	0.32	0.19	0.19
moq-tokio::integration	receive_credit::lite_frame_of_many_windows_completes	0.57	0.32	0.36	0.31
moq-mux	jitter::tests::start_up_latency_never_fails_a_full_multiplex	0.62	0.31	0.30	0.30
moq-mux	jitter::tests::a_source_past_the_drift_limit_fails	0.38	0.29	0.16	0.17
moq-stats	produce::tests::steady_drain_collects_without_allocating	0.44	0.29	0.28	0.28
moq-mux	container::ts::export_test::scte35_fixtures_survive_roundtrip	0.37	0.25	0.26	0.43
moq-net	model::track::test::an_empty_successor_still_lets_the_gaps_go	1.83	0.25	0.25	0.26
moq-net	model::track::test::an_endless_untimed_track_with_gaps_stays_bounded	1.82	0.24	0.23	0.24
moq-hls	export::tests::a_fresh_join_reads_a_bounded_number_of_records	0.67	0.21	0.21	0.20
moq-net::integration	compress::a_start_group_arrives_whole	0.35	0.20	0.20	0.19
moq-net	model::resume::test::a_resumed_reader_forgets_gaps_past_its_budget	1.46	0.12	0.12	0.18
moq-net	model::track::test::an_endless_track_with_gaps_stays_bounded	7.40	0.12	0.10	0.10
moq-mux	container::ts::import::test::a_signalled_reset_reaches_the_exported_clock	6.46	0.11	0.11	0.16
moq-mux	container::ts::import::test::a_shared_forward_boundary_resets_once_per_rendition	6.52	0.07	0.08	0.08
moq-net	model::track::test::an_untimed_sparse_feed_keeps_its_gaps	3.74	0.06	0.06	0.06
	Total	137.83	100.55	61.02	67.49

Merge history

• 36858e5 merges main 4b8e37b onto the land fixer's 31df580. Conflicts: main's moq_clock to kio::clock rename (moq-dev/moq#7027) in export_timing_test.rs (kept this PR's Deadlines harness, renamed) and in quest/m0/ci/slow-tests.md (kept this PR's deletion), and main's subscribe_at in moq-quic tests/util.rs (see Changes). five-minutes.md no longer waits on slow-tests.
• 38896af fixes the drain hang above; a merge of main 3164bce had no conflicts, and cf952d0 updates the Import::new doc for it.

Review

• Reviewer: Codex (quest review), gpt-6.1-sol, medium effort, session 01a19828.
• Round 1 on 36858e5: no findings. Round 2 on 38896af: no findings. Round 3 on cf952d0: no findings.
• Reviewed head: cf952d0.

(Written by Claude Opus 5.5)

🤖 Generated with Claude Code

[coderabbitai]

Warning

Review limit reached

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

Next included review available in 18 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used all 4 included reviews currently available.

Learn how review limits work.

Review configuration:

Run configuration

• Source mode: Auto
• Repository: Unavailable
• Branch: Unavailable
• File: No configuration file from this repository applied
• Commit: Unavailable
• Layers: Organization UI
• Review profile: CHILL
• Plan: Advanced
• Run ID: 5cdec293-fc36-417b-9cf6-01e5187ab851

Commits

Reviewing files that changed from the base of the PR and between 3164bce and cf952d0.

Files selected for processing: 33

• Cargo.toml
• quest/m0/ci/README.md
• quest/m0/ci/five-minutes.md
• quest/m0/ci/slow-tests.md
• rs/moq-audio/src/decode/consumer.rs
• rs/moq-cli/src/publish.rs
• rs/moq-hls/src/import.rs
• rs/moq-json/src/snapshot/encoder.rs
• rs/moq-json/src/stream/encoder.rs
• rs/moq-json/src/stream/mod.rs
• rs/moq-json/src/stream/producer.rs
• rs/moq-json/src/window/encoder.rs
• rs/moq-json/src/window/properties.rs
• rs/moq-mux/src/container/ts/export_timing_test.rs
• rs/moq-mux/src/container/ts/fixture.rs
• rs/moq-mux/src/container/ts/passthrough_test.rs
• rs/moq-mux/src/container/ts/recovery_test.rs
• rs/moq-mux/src/jitter.rs
• rs/moq-mux/src/rate.rs
• rs/moq-net/src/ietf/publisher.rs
• rs/moq-net/src/lite/publisher.rs
• rs/moq-net/src/model/front.rs
• rs/moq-net/tests/integration/routing_contracts/handoff.rs
• rs/moq-net/tests/integration/routing_contracts/model.rs
• rs/moq-pattern/tests/integration/pattern_exhaustive.rs
• rs/moq-quic/src/tests/probe.rs
• rs/moq-quic/src/tests/sim.rs
• rs/moq-quic/src/tests/sim/workloads.rs
• rs/moq-quic/src/tests/util.rs
• rs/moq-relay/tests/integration/cluster_unknown.rs
• rs/moq-tokio/tests/integration/broadcast.rs
• rs/moq-video/src/encode/encoder.rs
• rs/qmux/src/session/tests.rs

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

Share

• X
• Mastodon
• Reddit
• LinkedIn

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-30T15:26:51.904467Z	78a9aba	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[kixelated]

Automated review by review (OpenAI)

Reviewed commit: 12a66db0e24c52adcdda05fc63c6da7a3a011598.

No actionable bugs found. The production changes remain narrow: file-only HLS imports still enforce same-origin checks before fetching, JSON group budgets retain their production defaults across resets, and the single-track jitter shortcut skips only the depth comparison used to choose between tracks. The routing-model and deadline-driven test-harness changes also preserve the relevant ordering and checks on static inspection.

Direction: removing repeated allocation, logging, hashing and setup work is appropriate here. The documented shorter drift runs, smaller busy-stream workloads and different deterministic property-test samples are explicit coverage tradeoffs, not proof of identical coverage; keep the existing bounds and regression checks as these fixtures evolve.

Verification: static review of the changed code and relevant surroundings, including an independent routing-model/export-timing/fixture pass. No tests or benchmarks run; the reported CPU improvements, full-suite results and before/after output comparisons were not independently reproduced.

[kixelated]

Merged main into the branch (55fe56c) and resolved the conflicts by hand:

• moq-json (stream/encoder.rs, stream/mod.rs, window/encoder.rs): main dropped per-record compression (moq-dev/moq#7046). I kept main's version and re-applied this PR's configurable max_group_bytes on top, with the tests' 32 KiB BUDGET in place of MAX_CACHE_BYTES. The window header test lost its per-compression loop, as it did on main.
• moq-mux recovery_test.rs: I kept main's decoder()/parse() split from moq-dev/moq#7074 and moved this PR's framehash -hash murmur3 swap into it.
• moq-mux export_timing_test.rs: main's new an_unmarked_step_keeps_the_export_rendering called Live::run with .await, but this PR made run synchronous, so I removed the .await.
• moq-quic tests/util.rs: I kept main's client::Config signature with this PR's debug_span.

The full just check origin/main run passes, 10514 tests. On the first run, one moq-rtc test failed with a DTLS handshake timeout under load. It passed when run alone and on the rerun.

(Written by Claude Opus 5.5)

[moq-bot]

quest land removed this pull request from the queue: Conflicts with the base: quest/m0/ci/slow-tests.md, rs/moq-mux/src/container/ts/export_timing_test.rs, rs/moq-quic/src/tests/util.rs
.

[moq-bot]

codex / Codex land fixer: This maintainer pick already used its mechanical fi·x attempt. Run /quest-iterate to diagnose the repeated failure, then pick it again.

(written by codex / Codex)

[moq-bot]

quest land removed this pull request from the queue: Conflicts with the base: quest/m0/ci/slow-tests.md, rs/moq-mux/src/container/ts/export_timing_test.rs, rs/moq-quic/src/tests/util.rs
.

[moq-bot]

Codex land fixer: fixer crashed (exit 1); run /quest-iterate. Recommendation: run /quest-iterate before picking it again.

quest land check: [Errno 2] No such file or directory: '/var/lib/quest-land-moq/moq-dev-moq/fix-qrkoPz/.git/objects/maintenance.lock'

(written by Codex)

[moq-bot]

quest land removed this pull request from the queue: Conflicts with the base: quest/m0/ci/slow-tests.md, rs/moq-mux/src/container/ts/export_timing_test.rs, rs/moq-quic/src/tests/util.rs
.

[moq-bot]

codex / Codex land fixer: Base merge conflicts can be re·solved mechanically, but the merged candidate cannot pass tests: `timeout --verbose --kill-after=5s 45s cargo test --locked -p moq-audio -p moq-cli -p moq-hls -p moq-json. Recommendation: Review the export timing fixture drain/finish path with the base kio::clock migration; re·solve the runtime stall with coverage preserved, then rerun all twelve affected packages. Mechanical merge atte

		}
	}

	/// Run the clock to `at`, letting every leg read whenever it would have on its own.
	fn advance(&self, at: Instant, legs: &mut [&mut Leg]) {
		while let Some(wake) = legs
			.iter()
			.filter_map(|leg| leg.wake())
rg: rs/moq-mux/src/container/hang/audio.rs: IO error for operation on rs/moq-mux/src/container/hang/audio.rs: No such file or directory (os error 2)
   Compiling moq-msf v0.5.1 (/work/rs/moq-msf)
   Compiling ciborium-ll v0.2.2
   Compiling webm-iterable v0.7.1
   Compiling ciborium v0.2.2
   Compiling criterion v0.8.2
   Compiling idna_adapter v1.2.2
   Compiling idna v1.1.0
   Compiling url v2.5.8

exec
/bin/bash -lc 'timeout --verbose --kill-after=5s 45s cargo test --locked -p moq-audio -p moq-cli -p moq-hls -p moq-json -p moq-mux -p moq-net -p moq-pattern -p moq-quic -p moq-relay -p moq-tokio -p moq-video -p qmux@0.7.0 container::ts::export_timing_test::the_video_held_at_the_end_goes_out -- --exact --nocapture > /tmp/quest-6987-hang.log 2>&1' in /work
 exited 124 in 30023ms:
exec
/bin/bash -lc 'tail -12 /tmp/quest-6987-isolated.log; tail -12 /tmp/quest-6987-hang.log; ps -eo pid,ppid,comm,etime,pcpu --sort=-pcpu | head -7' in /work
 succeeded in 0ms:
   Compiling idna_adapter v1.2.2
   Compiling idna v1.1.0
   Compiling url v2.5.8
   Compiling bench-harness v0.0.1 (/work/rs/bench-harness)
   Compiling kio v0.6.2 (/work/rs/kio)
   Compiling mp4-atom v0.17.0
   Compiling moq-net v0.3.10 (/work/rs/moq-net)
   Compiling moq-json v0.5.8 (/work/rs/moq-json)
   Compiling moq-loc v0.2.19 (/work/rs/moq-loc)
   Compiling hang v0.21.11 (/work/rs/hang)
   Compiling moq-mux v0.10.11 (/work/rs/moq-mux)
timeout: sending signal TERM to command ‘cargo’

test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 3 filtered out; finished in 0.00s

     Running unittests examples/teleop.rs (target/debug/examples/teleop-bb6260ca7764be9c)

running 0 tests

test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 1 filtered out; finished in 0.00s

     Running unittests src/lib.rs (target/debug/deps/moq_mux-2113f186acb243cc)

running 1 test
    PID    PPID COMMAND             ELAPSED %CPU
  37403   37376 moq_mux-2113f18       00:02  3.1
  37376   37375 cargo                 00:35  1.7
     40      21 python3               08:39  0.5
     83      71 codex                 08:34  0.5
     71      40 quest                 08:34  0.0
     20       1 nsncd                 08:39  0.0

 exited 124 in 44882ms:
exec
/bin/bash -lc "tail -8 /tmp/quest-6987-hang.log; ps -eo pid,ppid,comm | rg 'cargo|moq_mux'" in /work
 exited 1 in 0ms:
running 0 tests

test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 1 filtered out; finished in 0.00s

     Running unittests src/lib.rs (target/debug/deps/moq_mux-2113f186acb243cc)

running 1 test
timeout: sending signal TERM to command ‘cargo’

codex
The filtered rerun reached `the_video_held_at_the_end_goes_out` and timed out with exit 124. I’ll hand PR 6987 back for review of the exporter-drain test fixture, restore its pinned branch, and record the failing command in the result file.

exec
/bin/bash -lc "git merge --abort
python3 - <<'PY'
import json
from pathlib import Path
cmd='timeout --verbose --kill-after=5s 45s cargo test --locked -p moq-audio -p moq-cli -p moq-hls -p moq-json -p moq-mux -p moq-net -p moq-pattern -p moq-quic -p moq-relay -p moq-tokio -p moq-video -p qmux@0.7.0 container::ts::export_timing_test::the_video_held_at_the_end_goes_out -- --exact --nocapture'
result=[{'number':6987,'action':'handback','diagnosis':f'Base merge conflicts can be resolved mechanically, but the merged candidate cannot pass tests: "'`{cmd}` reaches `running 1 test` and exits 124 with "timeout: sending signal TERM to command ‘cargo’"; the full package test run also reports `test container::ts::export_timing_test::the_video_held_at_the_end_goes_out has been running for over 60 seconds` (four other export timing tests also stall).'"','recommendation':'Review the export timing fixture drain/finish path with the base kio::clock migration; resolve the runtime stall with coverage preserved, then rerun all twelve affected packages. Mechanical merge attempt was aborted; pr-6987 remains at its pinned head.'}]
Path('quest-land-result.json').write_text(json.dumps(result,indent=2)+'\\n')
PY
 git status --short
 git rev-parse pr-6987 batch base" in /work
 succeeded in 2595ms:
Updating files:  47% (623/1321)
Updating files:  48% (635/1321)
Updating files:  49% (648/1321)
Updating files:  50% (661/1321)
Updating files:  51% (674/1321)
Updating files:  52% (687/1321)
Updating files:  53% (701/1321)
Updating files:  54% (714/1321)
Updating files:  55% (727/1321)
Updating files:  56% (740/1321)
Updating files:  57% (753/1321)
Updating files:  58% (767/1321)
Updating files:  59% (780/1321)
Updating files:  60% (793/1321)
Updating files:  61% (806/1321)
Updating files:  62% (820/1321)
Updating files:  63% (833/1321)
Updating files:  64% (846/1321)
Updating files:  65% (859/1321)
Updating files:  66% (872/1321)
Updating files:  67% (886/1321)
Updating files:  68% (899/1321)
Updating files:  69% (912/1321)
Updating files:  70% (925/1321)
Updating files:  71% (938/1321)
Updating files:  72% (952/1321)
Updating files:  73% (965/1321)
Updating files:  74% (978/1321)
Updating files:  75% (991/1321)
Updating files:  76% (1004/1321)
Updating files:  77% (1018/1321)
Updating files:  78% (1031/1321)
Updating files:  79% (1044/1321)
Updating files:  80% (1057/1321)
Updating files:  81% (1071/1321)
Updating files:  82% (1084/1321)
Updating files:  83% (1097/1321)
Updating files:  84% (1110/1321)
Updating files:  85% (1123/1321)
Updating files:  86% (1137/1321)
Updating files:  87% (1150/1321)
Updating files:  88% (1163/1321)
Updating files:  89% (1176/1321)
Updating files:  90% (1189/1321)
Updating files:  91% (1203/1321)
Updating files:  92% (1216/1321)
Updating files:  93% (1229/1321)
Updating files:  94% (1242/1321)
Updating files:  95% (1255/1321)
Updating files:  96% (1269/1321)
Updating files:  97% (1282/1321)
Updating files:  98% (1295/1321)
Updating files:  99% (1308/1321)
Updating files: 100% (1321/1321)
Updating files: 100% (1321/1321), done.
?? quest-land-result.json
31df5805b61560b12c84f2649269427ba2d1b4cd
31df5805b61560b12c84f2649269427ba2d1b4cd
4b8e37bd44e407b3416a17992720870ea3f9221c

codex
PR 6987 handed back: the merged candidate builds, but five export timing tests stall; a filtered rerun timed out.

Branch restored unchanged. Diagnosis and command recorded in uncommitted [quest-land-result.json](/work/quest-land-result.json).
tokens used
113,682
PR 6987 handed back: the merged candidate builds, but five export timing tests stall; a filtered rerun timed out.

Branch restored unchanged. Diagnosis and command recorded in uncommitted [quest-land-result.json](/work/quest-land-result.json).
fixer result written

(written by codex / Codex)

[kixelated]

Automated review by review (OpenAI)

Reviewed commit: cf952d0c25f0baff1feb51251eb8dabe27bdd629.

No new actionable bugs found in the update since review #5700676453, accounting for the merged base changes. The export-timing fixture now drives its own deadline queue instead of awaiting a timeout on that unadvanced queue. Its final-video count and paired-output assertions remain intact. This addresses the fixture stall described in the land-fixer report; it was not a finding in my previous review.

Direction: keeping the drain synchronous on the harness-owned clock is simpler and appropriate. The merged JSON per-record limit, QUIC logging/buffer adaptations, and HLS file-only path retain the intended narrow optimization scope; no additional public API or wire change is introduced by these updates.

Verification: static inspection of the incremental changes and relevant exporter, clock, and HLS code only. No tests or benchmarks run, and the reported gate results and CPU savings were not independently reproduced.

(Written by OpenAI)